English

Sample distribution theory using Coarea Formula

Probability 2021-12-06 v2

Abstract

Let (Ω,Σ,p)\left(\Omega,\Sigma,p\right) be a probability measure space and let X:ΩRkX:\Omega\to{\mathbb{R}}^k be a (vector valued) random variable. We suppose that the probability pXp_X induced by XX is absolutely continuous with respect to the Lebesgue measure on Rk{\mathbb{R}}^k and set fXf_X as its density function. Let ϕ:RkRn\phi:{\mathbb{R}}^k\to {\mathbb{R}}^n be a C1C^1-map and let us consider the new random variable Y=ϕ(X):ΩRnY=\phi(X):\Omega\to{\mathbb{R}}^n. Setting m:=max{\mboxrank(Jϕ(x)):xRk}m:=\max\{\mbox{rank }(J\phi(x)):x\in{\mathbb{R}}^k\}, we prove that the probability pYp_Y induced by YY has a density function fYf_Y with respect to the Hausdorff measure Hm{\mathcal{H}}^m on ϕ(Rk)\phi({\mathbb{R}}^k) which satisfies \begin{align*} f_Y(y)= \int_{\phi^{-1}(y)}f_X(x)\frac{1}{J_m\phi(x)}\,d{\mathcal{H}}^{k-m}(x), &\quad \text{for Hm{\mathcal{H}}^m-a.e.}\quad y\in\phi({\mathbb{R}}^k). \end{align*} Here JmϕJ_m\phi is the mm-dimensional Jacobian of ϕ\phi. When JϕJ\phi has maximum rank we allow the map ϕ\phi to be only locally Lipschitz. We also consider the case of XX having probability concentrated on some mm-dimensional sub-manifold ERkE\subseteq{\mathbb{R}}^k and provide, besides, several examples including algebra of random variables, order statistics, degenerate normal distributions, Chi-squared and "Student's t" distributions.

Keywords

Cite

@article{arxiv.2110.01441,
  title  = {Sample distribution theory using Coarea Formula},
  author = {Luigi Negro},
  journal= {arXiv preprint arXiv:2110.01441},
  year   = {2021}
}

Comments

22 pages

R2 v1 2026-06-24T06:36:25.256Z